KRAIL: A knowledge-driven framework for human reliability analysis integrating IDHEAS-DATA and large language models

Human reliability analysis (HRA) is crucial for evaluating and improving the safety of complex systems. Recent efforts have focused on estimating human error probability (HEP), but existing methods often rely heavily on expert knowledge, which can be subjective and time-consuming. Inspired by the success of large language models (LLMs) in natural language processing, this paper introduces KRAIL, a novel two-stage framework for knowledge-driven reliability analysis, integrating IDHEAS-DATA, knowledge graph and LLMs. The knowledge graph serves as a retrieval-augmented generation (RAG) layer that swiftly surfaces context-relevant evidence, while an expert-in-the-loop validation step alleviates data sparsity and curbs LLM hallucinations. Comprehensive experiments on authoritative HRA benchmark datasets show that KRAIL produces more accurate HEP estimates than state-of-the-art methods, even under partial-information conditions, while completing end-to-end assessments in under 150 s. These results underscore KRAIL’s potential to enable fast, transparent, and scalable human-error quantification for risk-informed decision making.